HiVA co-evolves agent prompts and graph topology from a single LLM agent using textual gradients from environment feedback, reporting consistent accuracy improvements over static and reactive agent baselines.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
HiVA: Self-organized Hierarchical Variable Agent via Goal-driven Semantic-Topological Evolution
HiVA co-evolves agent prompts and graph topology from a single LLM agent using textual gradients from environment feedback, reporting consistent accuracy improvements over static and reactive agent baselines.